Change np.core.defchararray to np.char (#9165)

[pont-us]

Replace a reference to np.core.defchararray with np.char.chararray in xarray.testing.assertions, since the former no longer works on NumPy 2.0.0 and the latter is the "preferred alias" according to NumPy docs. See Issue #9165.

• Closes testing.assert_allclose on string dtype gives error with NumPy 2.0.0 #9165
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst

[welcome]

Thank you for opening this pull request! It may take us a few days to respond here, so thank you for being patient.
If you have questions, some answers may be found in our contributing guidelines.

[keewis]

Thanks for the fix, @pont-us.

Could you add a test to make sure we don't regress, please? This could be as simple as adding another parameter set to

xarray/xarray/tests/test_assertions.py

Lines 39 to 61 in 872c1c5

	@pytest.mark.parametrize(
	"obj1,obj2",
	(
	pytest.param(
	xr.Variable("x", [1e-17, 2]), xr.Variable("x", [0, 3]), id="Variable"
	),
	pytest.param(
	xr.DataArray([1e-17, 2], dims="x"),
	xr.DataArray([0, 3], dims="x"),
	id="DataArray",
	),
	pytest.param(
	xr.Dataset({"a": ("x", [1e-17, 2]), "b": ("y", [-2e-18, 2])}),
	xr.Dataset({"a": ("x", [0, 2]), "b": ("y", [0, 1])}),
	id="Dataset",
	),
	),
	)
	def test_assert_allclose(obj1, obj2) -> None:
	with pytest.raises(AssertionError):
	xr.testing.assert_allclose(obj1, obj2)
	with pytest.raises(AssertionError):
	xr.testing.assert_allclose(obj1, obj2, check_dim_order=False)

or

xarray/xarray/tests/test_assertions.py

Lines 33 to 36 in 872c1c5

	def test_allclose_regression() -> None:
	x = xr.DataArray(1.01)
	y = xr.DataArray(1.02)
	xr.testing.assert_allclose(x, y, atol=0.01)

You might also want to add a note to whats-new.rst

[pont-us]

You might also want to add a note to whats-new.rst

I didn't think this would qualify as a "notable" bug fix, but glad to oblige :)

[keewis]

looks good to me, just one small nit.

[pont-us]

From a quick look, it appears that the failing Doctests are due to other, unrelated NumPy 2 changes (mainly NEP 51).

[keewis]

yep, this is also fails on main, so you can ignore these.

[welcome]

Congratulations on completing your first pull request! Welcome to Xarray! We are proud of you, and hope to see you again!

[max-sixty]

Thanks @pont-us

(any other changes for numpy 2.0 would be greatly appreciated if you're up for it...)

[keewis]

just FYI, the doctest / rtd will be fixed by #9177, and the normal CI will be fixed by #9136 (we just need to figure out whether using view is a good idea). This leaves us with the pydap plugin test, but since pydap raises an error with numpy>=2 on import there's nothing we can do about this other than disabling in CI and raising an issue on the pydap issue tracker.

[max-sixty]

That's awesome, thanks @keewis !

[pont-us]

Thanks for the help @keewis and thanks for the thanks @max-sixty :). I don't think I'll have time for another PR in the short term, unfortunately, but I plan to at least open issues for one or two other NumPy-2-related surprises that I've come across (if nobody else does first).